Drugs of the Future - Bispecific Antibodies

(1)

19-X2

Drugs of the Future - Bispecific Antibodies

An investigation of future development needs

Thomas Andersson, Anisha Khan, Therese Koivula, Terese Larsson,

Fabian Svahn, Amanda Wahlsten

Client: GE Healthcare Bio-Sciences AB

Client representative: Daniel Larsson

Supervisor: Karin Stensjö

1MB332, Independent Project in Molecular Biotechnology, 15 hp, spring semester 2019 Master Programme in Molecular Biotechnology Engineering

Biology Education Centre, Uppsala University

(2)

Abstract

This report reviews the field of bispecific antibodies, artificially engineered antibodies that have the ability to bind two or more different antigen simultaneously. Historical as well as recently developed techniques are demonstrated, together with formats in preclinical and clinical development. We studied the field with the future needs of the developers in mind, when it comes to the processes and tools that can be offered by GE Healthcare Biosciences AB.

The development of bispecific antibodies gave rise to new challenges and product-related impurities, which are handled by various methods. We argue for, based on the formats in clinical and preclinical development, that the methods already used to purify monospecific antibodies remain the most successful methods for the purification of bispecific antibodies.

This, together with the design strategies that resolve the initial bottle-necks, ensures that the needs of the developers are met to the same extent as for monoclonal antibodies. The methods and formats demonstrated here do not represent all that are available or under trial.

(3)

Abbreviations

BEAT - bispecific engagement by antibodies based on the T cell receptor BiTE - bispecific T cell engager

bsAb - bispecific antibody CH - constant heavy chain

CHO cell - chinese hamster ovary cell CL - constant light chain

DVD-Ig - dual-variable domain immunoglobulin Fab - antigen-binding fragment

Fc - crystallizable fragment/constant region FcR - Fc receptor

Fv - variable fragment HC - heavy chain

HEK cell - human embryonic kidney cell Ig - immunoglobulin

IgG/IgA/IgD/IgE/IgM - different classes of immunoglobulins present in humans KIH - knob-into-hole

LC - light chain

mAb - monoclonal antibody

MAT-Fab - monovalent asymmetric tandem Fab bispecific antibodies scFv - single chain variable fragment

SEEDBody - strand-exchange engineered domain SEC - size exclusion chromatography

VH - variable heavy chain VL - variable light chain

(4)

Definitions

Asymmetric design - antibody design that has no plane of reflection.

Bispecific - an antibody with the ability to bind two different antigens.

Bivalent - when an antibody binds only two antigens.

Design strategy - a way to design antibodies to circumvent product-related impurities.

Downstream - the processes involved in the harvesting and purification of a certain product.

Epitope - the part of the antigen that is recognized by the antibody.

Format - the structure of an antibody, unrelated to the choice of antigen binding sites.

Heterodimer - an antibody formed by association of two different polypeptides.

Homodimer - an antibody formed by association of two of the same polypeptides.

Monospecific - an antibody able to bind only type of antigen.

Monovalent - an antibody able to bind only one antigen at a time.

Immunogenicity - the ability of a substance such as an antigen or epitope to induce an immune response.

Parental antibodies - antibodies used to produce e.g. a bispecific antibody.

Platform - a method to obtain a certain antibody format.

Single-gene constructs - a product produced by transcription from a single gene.

Symmetric design - antibody design made up of identical parts facing each other.

Upstream - the processes involved in the design and expression of a certain product.

How we refer

If a reference is placed at the end of a sentence before full stop, it refers only to the sentence itself. If a reference is placed after a full stop, it refers to all the text in the paragraph between this reference and any previous reference.

(5)

1 Introduction

The market for recombinant proteins has in the later years been dominated by monoclonal antibodies (mAbs). Monoclonal antibodies are monospecific, meaning they target one type of antigen. These artificial antibodies have been successful as drugs in e.g. oncology.

Today, a new type of molecule with great therapeutic potential is on the rise. Bispecific antibodies (bsAbs) are artificially engineered antibodies that have the ability to bind two or more antigen simultaneously. An increasing number of bispecific antibodies enter clinical trials and there are already over 100 different formats of bsAbs (Brinkmann & Kontermann 2017). Bispecific antibodies are thereby opening up for entirely new therapeutic applications, as well as improved old ones.

GE Healthcare Biosciences AB (GE Healthcare) suggested this project because they are involved in the purification of monoclonal antibodies and therefore want to know the future needs of their clients. The project gives an overview of the field of bispecific antibodies where we have explored problems and their solutions, and listed formats in clinical and preclinical development. When looking at the different formats we have studied their structure, therapeutic effect and upstream and downstream processes. We have also listed advantages and disadvantages. We consider these topics to be the most important for understanding the field. The formats in clinical development demonstrate the varying structures and design strategies currently used, while the formats in preclinical development demonstrate what will become relevant for GE Healthcare in the future. An overview of the formats and strategies demonstrated in this report can be seen in figure 1.

Figure 1. A schematic figure of the clinical and preclinical bispecific antibody formats and design strategies described in this report. For further details about strategies see section 3.1 and 3.2. For further details about clinical and preclinical formats see chapter 4 and 5, respectively.

(8)

2

Based on our investigation of the formats in clinical and preclinical development, one of our conclusions is that the methods already used to purify monospecific antibodies remain the most successful methods for the purification of bispecific antibodies. The production of bispecific antibodies initially suffered from a low yield and product-related impurities.

However, many different innovative strategies combat this issue. Taken together, we conclude that the downstream production of bispecific antibodies no longer differs significantly from monoclonal antibodies, and that the needs of GE Healthcare’s potential clients are met. A

“business-as-usual” approach is therefore viable. This conclusion is described in more detail in chapter 6.

2 Understanding the structure and production of

antibodies

This chapter explains the structure and production of antibodies, both the naturally occurring and their engineered bispecific relatives. The development of bispecific antibodies is based upon the existing field of mAbs. It is therefore important to demonstrate these mAbs, before moving on to bispecific antibodies.

2.1 Monoclonal antibodies

Native antibodies or immunoglobulins (Ig) are glycoproteins produced by one type of immune cells called B-lymphocytes, as a part of the adaptive immune system. The mAbs are natural or recombinant antibodies that are produced by hybridoma cells. Hybridoma cells are formed by fusion of B cells and cancerous plasma cells called myeloma cells. All monoclonal antibodies produced by one hybridoma cell line are thus identical. (Wood 2011) Monoclonal antibodies are usually bivalent i.e. can bind two epitopes. An epitope is the part of the antigen which an antibody targets. The mAbs are monospecific since it can only bind to one type of antigen, i.e. the two antigen binding sites are identical. (Lodish 2016)

The most common isotype of antibodies produced by the immune system in humans is

immunoglobulin G (IgG). It consists of four different polypeptide chains, see figure 2. Two of these are identical heavy chains (HC). The heavy chains both consist of one variable domain denoted VH and three constant domains denoted CH1, CH2 and CH3 respectively. The C- terminal of the heavy chains is at the end of the CH3 domain. (Lodish 2016) The heavy chains are held together at the hinge region which is responsible for a covalent linkage between the two chains. It is moreover a flexible region which enables the distance between the Fabs to vary in length. (Brinkmann & Kontermann 2017) There are also two identical light chains (LC) which are covalently attached to the heavy chains by disulfide bonds. These consist of one variable domain denoted VL and one constant domain denoted CL. The domains of bsAbs are denoted in the same way. The light chains together with the VH and CH1 forms the antigen binding fragments (Fabs) that can bind a single antigen. (Lodish 2016)

The "spine" of the antibody that covalently links the two heavy chains is called the

crystallizable fragment (Fc). It consists of the CH3 and CH2 domains of both heavy chains

(9)

3

and can bind various receptors. These Fc receptors (FcRs) give the Fc region the ability to enable different effector functions in the immune system. Antibodies with Fc regions can bind to Fc receptors of certain immune cells that mediate cytotoxicity i.e. the killing of cells. This cytotoxic activity can cause high immunogenicity. (Wood 2011) We have seen that, in an attempt to reduce this, some formats in development utilize modified IgGs with reduced effector functionality. The Fc region also binds the neonatal Fc receptor which significantly extends the half-life of IgG antibodies (Kuo & Aveson 2011). A longer half-life of a

therapeutic antibody is associated with many benefits with some being improved treatment efficiency, fewer treatments and reduced costs.

Figure 2. A schematic figure of an IgG mAb. The IgG mAb consists of two pairs of heavy (HC) and light chains (LC) bound together by disulfide bridges shown as black bars. The two heavy chains are displayed in blue while the two light chains are shown in green. One HC consists of the domains VH, CH1, CH2 and CH3. One LC consists of the domains VL and CL. The two Fab regions constitute the two binding regions of the antibody.

The crystallizable fragment (Fc), also known as the constant region, enables the antibody to bind different receptors.

Besides the IgG isotype, there are four other isotypes of Fc regions in mammals; IgM, IgA, IgD and IgE. The Ig gets its name and class based on its Fc isotype. These have different functions in the immune response. The light chain also has two isotypes; kappa (κ) and lambda (λ). These are characterized based on the amino acid sequence in the constant region.

A mAb will contain two heavy chain associated with either two κ chains or two λ chains.

(Lodish 2016)

2.2.1 Today’s purification methods of monoclonal

antibodies

According to GE Healthcare’s handbook “Affinity chromatography, Vol.1: Antibodies”, the purification of mAbs is most often done using affinity chromatography. If high purity is required, a further polishing step can be performed. For this, size exclusion chromatography

(10)

4

(SEC) is frequently used. The chromatography resins, i.e. the stationary phase, contains an immobilized ligand which has affinity to the mAb of interest. This gives affinity

chromatography its high specificity. There are, according to the handbook, three commonly used affinity ligands: protein A, protein G and protein L. Protein A and G both bind to the Fc region of the antibody while protein L binds to the VL domain of the κ isotype. The

interaction with protein L does not require any Fc region which gives it a wider range of possible target antibodies. For more details on available purification methods, see appendix 1.

2.3 Bispecific antibodies

The main difference between mAbs and bispecific antibodies is that bsAbs can bind two different epitopes, either on the same or another antigen, see figure 3. They are, excluding IgG4, artificial proteins not found in nature. Currently, bispecific antibodies are commonly produced via fusion in mammalian cell lines. The most prevalent cell lines are CHO and HEK cells. For more information about these cell lines and how they are used to produce

antibodies, see appendix 2.

Figure 3. A schematic figure of a bispecific antibody. The chains displayed in pink and purple are the heavy and light chains from one monospecific antibody. The heavy and light chain of the other monospecific antibody are displayed in blue and green, respectively.

There are nowadays more than 100 different formats of bsAbs. Today, an increasing number of bsAb enter clinical trials and some are already on the market. (Brinkmann & Kontermann 2017) As it will be apparent in this report, the structure of different formats varies a lot. From small formats with only variable regions, i.e. fragment-based/non-IgG-like formats, to larger formats with constant regions, i.e. Fc-based/IgG-like formats (Fan et al. 2015). There is no single “best format”. Different formats come with varying advantages and disadvantages and can be successful for different therapies (Brinkmann & Kontermann 2017).

There are various applications of bsAbs including diagnosis, prophylaxis and therapy. A common approach is the retargeting of effector cells, such as T cells, to a tumor cell for

(11)

5

cancer therapy. This is possible by designing the bsAbs to bind one epitope on the T cell and one epitope on the tumor cell and in this way directing the immune system to the diseased cell. (Duell et al. 2019) Bispecific antibodies are also potential pharmaceuticals for treatment of other diseases such as autoimmunity, bleeding disorders and infections. (Brinkmann &

Kontermann 2017)

2.3.1 Fragments and linkers are common structural elements

As mentioned, there are many formats of bsAbs which may vary a lot in structure. Different fragments are occurring as parts of bsAbs. Some frequently used fragments can be seen in figure 4. As can be seen in the figure, some of the fragments are connected by linkers. We have seen that the most often used linker is the G4S-linker. Read more about linkers in appendix 3.

Figure 4. Some frequently used fragments in bsAbs as presented in “Antibody Production in Microbial Hosts”

by Rathore & Batra (2016). scFv: single chain variable fragment. A VH and a VL domain connected by a linker.

Fab: antigen binding fragment. Consists of HC constant and variable region connected with LC constant and variable region. scFab: single chain antigen binding fragment. A Fab with a linker between the HC and LC.

Diabody: consists of two VH/VL domains connected by linkers. sVD: single variable domain comprised of a VH, often derived from camelid or shark antibodies.

2.3.2 The quadroma method pioneered the construction of

bispecific antibodies but suffers from low yield

The quadroma method was one of the first developed methods for producing bispecific antibodies. The method requires two hybridomas that are combined by somatic fusion generating a hybrid hybridoma, also called a quadroma. The hybrid cell line in a quadroma co-expresses the genes encoding two heavy chains and two light chains. Each pair of heavy chain and light chain are from respective parental hybridomas. (Schaefer et al. 2015) This allows the combination of heavy and light chains of two different monoclonal antibodies in one single cell. The constant region of these chains can be of the same or different isotypes

(12)

6

and even retrieved from different species. (Brinkmann & Kontermann 2017) The resulting bsAbs generated from quadroma cell lines have the binding properties of their parental antibodies and resemble conventional IgG-like antibodies (Fan et al. 2015). Quadroma

continues to be used in the development of novel bispecific antibody formats. One of the only two bispecific antibodies that have received market approval is catumaxomab (Creative Biolabs, 2019). Catumaxomab is a trivalent Triomab produced by the quadroma technology where cell lines from rat and mouse are utilized (Duell et al. 2019).

However, there are issues with the quadroma technology. One of the main problems is the formation of unwanted antibodies along with the target bsAbs, see figure 5. There can be up to nine variants of unwanted antibodies that are either non-functional or monospecific. These variants may also be referred to as product-related impurities. The excess formations are due to homodimerization instead of heterodimerization of the heavy chains as well as random binding of light chains with heavy chains. The former is referred to as the heavy chain problem while the latter is referred to as light chain mispairing. (Sedykh et al. 2018) Homodimerization refers to when two heavy chains with the same binding properties are associated meanwhile heterodimerization is the opposite, hence resulting in a bispecific antibody. Light chain mispairing, on the other hand, is the incorrect association of a certain heavy and light chain pair. This results in non-functional binding domains for the specific antigen the bsAb is being constructed for. (Brinkmann & Kontermann 2017) This normally leads to a low yield of target bsAbs which is a significant disadvantage with the quadroma technology (Sedykh et al. 2018).

Figure 5. A schematic figure of the possible heavy and light chain formations in a quadroma cell line. The green and blue chain represent the light and heavy chain from one of the parental antibodies. The purple and pink chains represent the light and heavy chain from the other parental antibody. The first two formations seen in the

“incorrect” group are biologically functional but not bispecific. The rest of the incorrect formations are not functional.

(13)

7

2.3.3 The transition from the quadroma method to recombinant

fusion constructs expressed in mammalian cell lines

All the formats studied and demonstrated in this report were the result of genetic fusion and expressed in mammalian cell lines. We interpret this to be due to the product-related

impurities when using the quadroma method. Expressing the different chains in a mammal cell does not solve the problem with product-related impurities. However, it allows for the expression of genetically modified chains. This, in turn, enables the use of different design strategies that combat the problem with incorrect chain pairing. Different design strategies are demonstrated in chapter 3, together with formats in clinical development. For information about the cell lines used in the production, see appendix 2.

3 An overview of design strategies

This chapter describes different strategies and approaches that solves issues regarding the manufacturing of bispecific antibodies. We have defined a strategy as a manufacturing approach that is applicable in the design of bsAbs where the format in development is not uniquely defined by the method used. A format will in this report refer to a structurally distinct antibody, not restricted to having a certain set of binding domains. In other words, two different formats can use the same strategy to solve a certain problem, while still being structurally different. The strategies described are presented in table 1, where whether they solve the heavy chain problem and/or light chain mispairing is indicated.

Table 1: Table showing the different strategies discussed in the report. It also includes what type of chain problem each strategy solves i.e. heavy chain (HC) problem and/or light chain (LC) mispairing.

Strategy Solves HC problem Solves LC mispairing

KIH Yes No

SEEDBody Yes No

BEAT Yes Yes

CrossMab No Yes

Orthogonal Fab No Yes

DuetMab No Yes

CH1-CL interface mutations No Yes

3.1 Strategies for solving the heavy chain problem

One way of diminishing the amount of product-related impurities is by solving the heavy chain problem. That is, promoting heterodimerization by making homodimerization

unfavorable. Three strategies that solve this problem, the pioneering knob-into-hole as well as SEEDBody and BEAT, are discussed below.

(14)

8

3.1.1 Knob-into-hole (KIH)

The KIH method is described in the patent “Knobs and holes heteromeric polypeptides”

(US8216805B2, 2012). The method utilizes the introduction of protuberances; smaller protruding elements, or “knobs”, at one or multiple interfaces of an initial polypeptide (US8216805B2, 2012; Ridgway et al. 1996). These protuberances are introduced together with corresponding cavities, or “holes”, at corresponding interfaces of an additional

polypeptide, see figure 6 (US8216805B2, 2012; Ridgway et al. 1996). Both the protuberances and cavities is introduced into two separate constructs of choice at their respective

corresponding interfaces, and is done by altering the nucleic acid encoding the polypeptide in question (Klein et al. 2016; US8216805B2, 2012). By replacing smaller amino acids side chains, e.g. alanine or threonine, with larger ones, e.g. tyrosine or tryptophan, it is possible to introduce protuberances at predetermined locations of interference (US8216805B2, 2012).

Corresponding compensatory cavities is subsequently possible to introduce into a polypeptide by means of replacing larger amino acid side chains with smaller ones.

The knobs and holes established are positioned in the CH3 region of antibodies in such a fashion as to hinder the formation of homodimers by making such formations sterically unfavorable. Subsequently, this promotes the formation of heterodimers, and does so by offering an effective and sterically favorably “docking system”, as can be seen in figure 6.

Figure 6. A schematic illustration of heterodimerization induced by the knob-into-hole method. The purple bar represents an engineered “knob” designed on one, here seen in red, antibody in such a way and place as to fit into a sterically complementary “hole” located on a, here blue, separate antibody monomer.

The knob- and hole-dimers may be expressed in a single cell line or in separate expression systems. The host cell can be transformed with a single vector or independent vectors

containing DNA encoding all polypeptides. After expression, the polypeptides are combined to form a fully assembled antibody. This allows the formation of desirable heterodimers with high precision and limited mispairing. However, this method is not limited by a set number of protuberance and cavities. Sterical unfavourability or favorability between heterodimers can

(15)

9

be increased or decreased by introducing multiple corresponding knobs and holes.

(US8216805B2, 2012)

3.1.2 SEEDBody

SEEDBody (Sb) is another approach to induce correct heavy chain pairing developed by Davis et al. (2010). SEED stands for strand-exchange engineered domain. Much like the KIH strategy, it manipulates the interface in the CH3 region to promote heterodimerization. The approach “interlocks” beta strand regions of the human IgG and IgA CH3 domains without introducing non-native disulfide bonds. Both SEED CH3 domains are built out of alternating CH3 beta strand sequences of human IgA and IgG. They are complementary, thus promoting correct heterodimerization, see figure 7. (Davis et al. 2010)

The yield of correctly heterodimerized Sb in general ranged from 85% to 95% of total protein purified from transfected human embryonic kidney (HEK) cells in a study by Davis et al.

(2010). The yield was also equivalent to Fc-based fusion proteins (Davis et al. 2010), known for their high expression in mammalian cell lines (Lo et al. 1998). For more information about mammalian cell lines, see appendix 2.

Figure 7. A schematic figure of the SEEDBody method. A: the two heavy chains with their engineered SEED CH3 domains depicted with red and blue bars. B: The two heavy chains associated as heterodimers. C: A schematic figure of the interlocking IgG and IgA beta strands at the CH3 domains.

The overlapping IgA and IgG sequences in the Fc region could potentially result in loss of important Fc functions, such as protein A or G affinity. Moreover, if binding to the neonatal Fc receptor is lost, the Sb would have significantly shorter half-life, since the neonatal Fc receptor prevents proteins from degradation. However, protein A binding was shown to be retained and the Fc region was also shown to confer longer half-life with the total half-life being equal to the control IgG antibody (Davis et al. 2010). They consider this to be an

“indirect but strong evidence that the Sb binds to the neonatal Fc receptor (FcRn) and is also

(16)

10

consistent with the observed retention of protein A association by the overlapping region of the Sb structure.”

This is good news for the SEEDBody approach, as the loss of FcRn binding would be a significant disadvantage compared with less disruptive methods. Davis et al. (2010) further suggest that fusing the Sb with a scFv to the N-terminal would be a good way of producing bsAbs since it evades the problem with light chain mispairing. We think that other methods to combat light chain mispairing described in this report could also be applied together with SEEDBody.

3.1.3 BEAT - Bispecific Engagement by Antibodies based on the T

cell receptor

BEAT was developed by the company Glenmark. Although the name BEAT technically only applies when a T cell receptor is used as binding site, we consider this a general strategy to solve the heavy chain problem. The second binding site has affinity to a generic disease associated antigen. In the leading BEAT format, the second target is the HER2 receptor on breast tumor cells. (Moretti et al. 2013; WO2015063339A1, 2015) This antibody is further described under clinical formats.

The BEAT strategy provides a way of producing full and correctly assembled heterodimeric immunoglobulins similar to native forms. The BEAT heavy chains each contain a modified CH3 domains to promote the formation of an interface in the Fc region. The engineered modifications are made by exchanging residues in the Ig constant domain of the antibody with residues from two domains in the T cell surface receptor. The T-cell receptor is partly formed by an interaction that creates a protein-protein interface between two domains called alpha and beta. This interaction is used in the BEAT technology to design new heterodimeric domains in the antibodies. Heterodimerization of the heavy chains therefore mimics the natural association of T cell receptors alpha and beta domain. This modification creates a stable interface which already occurs naturally and is applicable to all BEAT antibodies.

(Skegro et al. 2017)

The standard format of BEAT antibodies is a scFv x Fab format, i.e. one of the antigens binding Fab arms is replaced by a scFv, see figure 8. This design avoids any possible light chain mispairing that otherwise could occur. It also allows versatile combinations because it is not restricted to the use of common light or heavy chains. (Moretti et al. 2013;

WO2015063339A1, 2015)

(17)

11

Figure 8. A schematic figure of BEAT as presented in the patent WO2015063339A1, 2015. The antibody includes one scFv represented in purple and one Fab arm.

To solve impurities related to homodimerization, the BEAT antibody is engineered with an asymmetric binding to protein A. The heterodimer has reduced or eliminated binding to protein A, which allows for efficient one-step purification via protein A chromatography. This can be created using Fc heavy chains from IgG isotypes with weak or no ability to bind

protein A, e.g. IgG3, or by substitutions in the Fc binding region. (US9683053B2, 2017;

WO2015063339A1, 2015)

Protein A can also bind the VH chain from the human subclass called VH3. VH3 chains have demonstrated improved expression and stability over other heavy chain subclasses. Therefore, VH3 antibodies have been widely developed in the biotechnology industry.

(WO2015063339A1, 2015) When using antigen binding sites from a VH3 origin, the protein A binding sequence has to be substituted or located on the heavy chain that does bind to protein A in the Fc region. This way, the purification problem is solved by the significant difference in affinity for protein A between the homo- and heterodimers. Two homodimers are formed, one with no protein A affinity and one with a much stronger protein A affinity.

(WO2015063339A1, 2015)

This is a unique way of designing bispecific antibodies to avoid the bottle-neck relating to formation of product-related impurities. Similar approaches have not been seen in our investigation, where other design strategies primarily focus on promoting correct association of heavy and light chain. The BEAT strategy not only promotes correct heterodimerization but also enables efficient separation of possible formation of homodimers from the product.

As mentioned before, a problem with bsAbs containing an Fc region is the interaction with receptor expressing immune cells inducing an unwanted cytokine release response. This is solved in the BEAT technology by substitutions in the Fc binding region. It can be done by introducing suggested substitutions in the constant region of the first and/or second

(18)

12

polypeptide. This produces a constant region without Fc receptor binding ability.

(WO2015063339A1, 2015)

3.1.4 The strategies to solve heavy chain pairing are not optimal for

large scale manufacturing and can be improved

KIH, SEEDBody and BEAT suffer from the necessity of delivering at least two gene constructs that need to be expressed proportionally in the same or different cell lines.

Manufacturing could be more easily scaled up if the formats could be expressed using single gene constructs. We might see more formats developed via single gene constructs in the future. However, the design strategies that require co-delivering of at least two gene constructs are still powerful tools to ensure correct heterodimerization of the heavy chains.

When constructing the BEAT formats, they also engineered varying protein A affinity into the heavy chain constant regions, facilitating the purification process of the impurities that are still formed. We argue for that this is not an option that is exclusive to the use of BEAT and could be applied to the KIH and SEEDBody strategies as well. There is however a pending patent application (WO2015063339A1, 2015) for this strategy.

3.2 Strategies for solving the light chain problem

Since applying the technologies discussed above only solves the problem with

homodimerization of the heavy chains, further methods have to be applied to solve light chain mispairing. That is, two different LC can give rise to four different antibodies were only one is bispecific, see figure 9.

Figure 9. A schematic figure of the four possible antibodies which can be generated after solving the HC problem. The encircled antibody is bispecific whereas the others are non-functional.

Solving this problem is more difficult than the heavy chain pairing issue because of the

complex interactions within the Fabs (WO2018035084A1, 2018). One easy approach to avoid this problem is to use common light chains, though it would drastically reduce the possible antigen binding sites. Below, the strategies CrossMab, orthogonal Fab, DuetMab and LC- mutations that solve the light chain problem are discussed. Usually, a light chain pairing

(19)

13

method is combined with an approach for the HC problem to ensure correct chain pairing overall. The strategies described in this section cause the format in question to deviate from the native IgG structure in varying degrees and may require a lot of engineering.

3.2.1 CrossMab

CrossMab is one of the first technologies developed to solve the light chain mispairing problem. It has evolved to become one of the most used and validated techniques. The advantages of CrossMab is that it is compatible with standard upstream techniques and purification processes used during the production of conventional IgGs. Furthermore, it conserves the properties of normal IgGs in terms of, for instance, stability and glycosylation.

CrossMab enables the making of bivalent antibodies but also trivalent and tetravalent antibodies can be made. (Klein et al. 2019)

The CrossMab technology produces correctly formed antibodies by making crossovers, see figure 10. This promotes correct light chain formation. Together with the use of KIH for correct heavy chain pairing, both the light chain problem and the heavy chain problem are solved. The crossover is done by exchanging domains within one Fab arm of the antibody.

The exchange can occur between different domains within the antigen binding fragment. A CrossMab^Fab is constructed by, on one side, designing a new heavy chain consisting of a Fc and the Fab of the original light chain and using the original VH and CH1 domains as the new modified light chain. The crossover can also be applied on only a part of the Fab arm by exchanging the variable domains producing CrossMab^VH-VL or CrossMab^CH-CL by exchanging constant domains, see figure 10. The domain interchange makes the two binding arms

significantly different and light chain mispairing can no longer occur due to not being able to form heterodimerization interfaces between light and heavy chain regions. Furthermore, this crossover can be done without affecting the binding affinities to the antigens. (Brinkmann &

Kontermann 2017, Schaefer et al. 2011)

(20)

14

Figure 10. A schematic figure of different CrossMab crossovers as inspired by figure 1 in an article by Schaefer et al. (2011). A commercial mAb and three different CrossMab versions resulting from different crossovers in the same Fab arm are depicted. With the CrossMab technology, the light chain mispairing problem is solved. For CrossMab^Fabthe VH and VL domains are swapped as well as the CL and CH1 domains. For CrossMab^VH-VL the VL and VH domains are swapped and for CrossMab^CH1-CLthe CL and CH1 domains are swapped. These swaps are represented by arrows. All three versions have the knob-into-hole modification represented as the purple bar.

Depending on the crossover domain, different side products can still be formed. This occurs especially when producing CrossMab^VH-VLor CrossMab^Fab. The side products are usually removed by applying different chromatography techniques. This can however be problematic and further solutions have been developed. For instance, to avoid formation of impurities when producing VH-VL CrossMabs, charged amino acid pairs can be introduced to promote correct assembly of the antibody by creating electrostatic attraction. (Brinkmann &

Kontermann 2017, Klein et al. 2019)

3.2.2 Orthogonal Fab

Orthogonal Fab solves the light chain problem by the genetic engineering of the interface between the heavy and light chains to get so called orthogonal Fab interfaces. This achieves higher affinity for matching light and heavy chains. When simultaneously expressing parental monoclonal antibodies incorporating these interfaces, heavy and light chain pairing in the obtained bispecific IgG antibody is improved with an average of 93% correctly assembled antibodies. (Lewis et al. 2014)

When engineering an orthogonal Fab it is important to take the number of mutations into account as it affects the expression of the antibody. In the study by Steven M Lewis et al.

(2014) a general trend was that “the designed proteins that were well expressed had fewer mutations, and all proteins with more than nine mutations were not expressed”. This was the

(21)

15

result from mutations in the CH1-CL and VH-VL interfaces.

This method has been successful when applied to antibodies with various specificities but does not give 100% correct assembly of the Fabs. It is worth mentioning that obtaining the correct assembly of Fabs was specific for the antibodies in the study and it is therefore unclear whether this strategy works for all antibodies. (Lewis et al. 2014)

3.2.3 DuetMab

DuetMab replaces the native disulfide bond in the CH1-CL interface with an engineered disulfide bond, see figure 11. (Mazor et al. 2015) This enhances cognate light chain pairing but needs to be combined with another design strategy that ensures correct heavy chain heterodimerization.

Figure 11. A schematic figure of the DuetMab approach. The orange line represents the engineered disulfide bond which replaces one of the native disulfide bonds.

Three different positions in the CH1-CL interface are possible candidates for favouring the formation of a novel disulfide bond according to the DuetMab approach. An amino acid on the HC and one on the LC is replaced with cysteine in one of the Fab regions. The native disulfide bond on the other Fab region is left intact. This approach was experimentally validated by Mazor et al. (2015) and applied together with KIH. Combined, they resulted in nearly 100% formation of bispecific antibodies.

It is advantageous that the modifications are in the CH1-CL interface and not in the variable domain, as this could have detrimental effects on antigen binding (Mazor et al. 2015).

Although, engineering in the CH1-CL interface could mean that κ and λ constant light chains would somehow affect the usefulness of this approach. However, it was shown to be

compatible to both isotypes (Mazor et al. 2015).

It could be that DuetMab cannot be applied to all Fab interfaces with such successful results.

Nonetheless, 32 other DuetMab antibodies were successfully developed by Mazor et al.

(2015), which points to the fact that the approach is generally applicable. Mazor et al. (2015)

(22)

16

concluded that DuetMab could be generically applied to bispecific antibodies in development since the approach: (i) does not contain variable domain engineering, (ii) is compatible with both kappa and lambda isotypes and (iii) was able to induce correct heterodimerization.

3.2.4 CH1-CL interface mutations

This design strategy is similar to the orthogonal Fab and DuetMab strategies in that it introduces modifications in the CH1-CL interface. The modifications disrupt pairing with wild type antibody chains and promote pairing with cognate engineered chains. (Bönisch et al. 2017)

This design strategy was developed by Bönisch et al. (2017). Their selected mutations were based on previously published repulsive CH3-CH3 interface mutations (Ying et al. 2012) due to structural similarity between the CH3-CH3 and CH1-CL interfaces. Modifications of the residues in the CH1-CL interface that were homologous to the repulsive CH3-CH3 were therefore explored and their impact on structure examined. When suitable candidates were found, complementary mutations to restore the CH1-CL interaction were examined to restore the association between the modified chains. The number of modifications negatively

correlated with expression which was also the case for Lewis et al. (2014) developing the orthogonal Fab approach.

Four different engineered interfaces survived experimental validation. The interfaces were able to promote cognate pairing with all interfaces by introducing mutations that repulsed the wild type chain together with mutations that restored the interaction with the modified chain.

Increases in yield were cell line dependent but still significant. A negative correlation between correct chain pairing and total yield was observed. The authors speculate that this could be due to that cell lines with high expression “overwhelm inherent quality control mechanisms”.

(Bönisch et al. 2017)

3.2.5 Excessive engineering in the light chain interfaces leads to

lower yield and can have implications for antigen binding affinity

The orthogonal Fab, CH1-CL interface mutations and the DuetMab approach involve varying degrees of engineering Fab in the interface. As was demonstrated by developers of the

orthogonal Fab and the CH1-CL interface mutations approach, the yield was negatively correlated to the amount of performed mutations (Mazor et al. 2015, Lewis et al. 2014).

Mutations in the variable region can interfere with the antigen binding affinity of the format, but not necessarily. The applicability of these strategies should vary depending on whether mutations are introduced in the variable region or not. The mutations introduced in CH1-CL interface mutations by Bönisch et al. (2017) might therefore be an approach that can be applied to a wider range of formats and targets than the orthogonal Fab strategy.

(23)

17

4 An overview of clinical formats

There are several bispecific antibodies in clinical trials today. Many of them use different formats, where each format has different properties. To get an insight of the formats that currently are being used in the field, a study of clinical formats was performed. For each respective format, the structures are described. Their therapeutic effects are also presented in order to understand where these molecules have potential as drugs. The upstream and

downstream process are described as well as the advantages and disadvantages of the formats.

A presentation of the bispecific antibodies in clinical trial described in this report can be seen in table 2.

Table 2. Overview of the clinical formats described including whether they contain a Fc region, possible purification strategies, advantages and disadvantages of the format. The possible purification strategies, and their respective abbreviations, are; PAC, protein A chromatography; PGC, protein G chromatography; PLC, protein L chromatography; SEC, size exclusion chromatography; IEC, ion exchange chromatography.

Name Fc region Possible purification

strategies

Advantages Disadvantages

DVD-Ig Yes PAC/PGC Avoids HC/LC

mispairing

Lower binding affinities

scFv fusions Yes Standard processes,

depending on conjugated protein

Longer half-life, avoids LC miss paring

Low stability because of linkers

BEAT Yes PAC/PGC No HC/LC

mispairing, avoids effects that might cause immunogenicity

Disallows the use of VH3 variable domains without further

engineering steps

XmAb Yes PAC, IEC Extended half-

life, purification advantages

Possible risk of immunogenicity

BiTE No His-tag, PLC High specificity Short half-life

TandAb No His-tag, PLC High specificity,

improved half- life over BiTE

Short half-life

(24)

18

4.1 Fc-based formats

As previously mentioned, formats of bispecific antibodies can primarily be grouped into Fc- based, i.e. IgG-like, and fragment-based, i.e. non-IgG-like. All formats described in this section belong to the IgG-like group, since they are based on the “native” form of an IgG antibody.

The Fc-based formats outperform the fragment-based in terms of half-life. However, the Fc- based formats have their own problems regarding immunogenicity and the formation of product-related impurities. Applying strategies to solve problems associated with chain pairing produces formats that in varying degrees differ from the native structure. Extensive alteration of the antibody structure, or the existence of remaining impurities, can lead to toxicity effects. (Wood 2011)

4.1.1 Dual-variable domains Ig (DVD-Ig)

Dual-variable domain immunoglobulin (DVD-Ig) is an established format. The company AbbVie have developed several DVD-Ig like bispecific antibodies that currently are in clinical trials. One product in clinical trial is ABT165. (Sedykh et al. 2018)

AbbVie have filed a patent application titled “Isolation And Purification Of DVD-Igs”

(US20160272673A1, 2016). The information described in the section “Production methods”

is based on the assumption that the methods and protocols presented in the application is what they apply to their DVD-Ig products. AbbVie have a granted patent “Anti-dll4/vegf Dual Variable Domain Immunoglobulin And Uses Thereof” (US9045551B2, 2015) for a DVD-Ig with the same bindings domains as ABT165. “Production methods” is also based on the assumption that the information in the patent can be applied to the DVD-Ig ABT165. The upstream and purification processes of this particular format have been supplemented from aforementioned patent application.

Structure

DVD-Ig is a format with an Fc region and bispecific characteristics on each Fab arm, making it a tetravalent bispecific antibody. This is achieved by additional fusions of a second variable domain to the variable domains of a monospecific antibody with a linker, see figure 12. (Fan et al. 2015)

(25)

19

Figure 12. A schematic picture of DVD-Ig as presented in figure A in the patent US9045551B2, 2015. The pink represents the additional binding domains fused to the monospecific antibody in blue and green.

Therapeutics effects

ABT165 targets solid cancer tumors by binding to the delta-like ligand 4 (DLL4) and VEGF receptor. (Sedykh et al. 2018) Both DLL4 and VEGF are angiogenic factors, they contribute to angiogenesis which is the process where new blood vessels is formed (Li et al. 2018).

When binding to these factors, the growth of the tumor is inhibited due to conformational changes in the ligand-receptor system, affecting the signaling to initiate angiogenesis. This is because the cells cannot survive without blood vessels. (Li et al. 2018; US9045551B2, 2015) Production methods

In the example in patent US9045551B2 (2015) the heavy and light chains that compose the variable domain for binding to DLL4 and VEGF was synthesized with two-step PCR. Known domain sequences from humans were used to design the chains. Primers were designed with flanking regions to the cloning vector as well as a linker region between each variable domain. These were inserted into a vector and positive cloning vectors were identified through bacterial transformation. After being harvested and purified, the vector encoding genes for the recombinant bispecific binding protein were expressed via mammalian host cells. (US9045551B2, 2015) Host cells may include CHO cells, NSO myeloma cells, COS cells and SP2 cells. (US20160272673A1, 2016)

Affinity chromatography is used for purification, preferably protein A chromatography. Here, the Fc region is utilized to capture the DVD-Ig. The patent (US20160272673A1, 2016) suggests that suitable resin for this matter is MabSelect or MabSelect SuRe from GE Healthcare or ProSep Ultra Plus from EMD Millipore. The purification of a DVD-Ig may include further steps of ion exchange chromatography and/or hydrophobic interaction chromatography as well. Such steps may be anion exchange chromatography, mixed mode chromatography of either cation exchange or anion exchange type, hydrophobic interaction chromatography and viral filtration. (US20160272673A1, 2016)

(26)

20 Advantages and disadvantages

One of the advantages with DVD-Ig is that it avoids the heavy chain problem as well as light chain mispairing since the Fab arms are identical. Therefore, no mispairing can occur which improves yield and the incorporation of an Fc region improves stability. (Fan et al. 2015) A possible disadvantage with the format is lower binding affinities to target molecules due to steric hindrance. This especially applies to the inner variable domain. The structure and length of the linker connecting two variable regions within one Fab arm also affect the binding affinity on the inner variable domain. To avoid this, the linker should preferably be longer, about 15 amino acids, and be adopting a helical structure. (Brinkmann & Kontermann 2017)

4.1.2 scFv fusions

This class of bispecific antibodies are generated by fusing fragments, or so-called binding motifs, to other protein domains. A review by Brinkmann & Kontermann displayed many applied formats of this kind with large diversity in their structure. For example, have a variety of binding moieties been applied, for example scFvs, Fabs, single chain Fabs (scFabs) or even single variable domains (sVD) derived from camelid and sharks, see figure 4. However, scFv are most commonly used among these. The protein domain can be a whole IgG, only a Fc region or alternatively other scaffold proteins. (Sedykh et al. 2018)

It is also possible to generate asymmetric bispecific antibodies using this approach for

example by fusing a Fab binding arm to a scFv-Fc moiety. This have been applied to formats that are further described later in this report. (Brinkmann & Kontermann 2017)

4.1.2.1 scFv - IgG fusions

One way of designing symmetric bispecific antibodies is by fusion of scFvs to an IgG (scFv- IgG). (Brinkmann & Kontermann 2017)

Structure

The fusion of the scFvs to the IgG can be done to the C-terminus of the heavy chain, or to the N-terminus of the heavy/light chain. The antibodies generated are therefore symmetric and usually tetravalent with two binding sites for each antigen. Thus, no heavy/light chain mispairing can occur in the production process. The first developed format of this kind was the Morrison format, see figure 13. Since then, many of the scFv-IgGs developed have been based on this format. The Morrison format was constructed by fusion of two scFvs targeting one antigen to the C-terminus of an IgG. (Brinkmann & Kontermann 2017, Sedykh et al.

2018)

(27)

21

Figure 13. A schematic picture of the Morrison format as presented in figure 3 in an article by Dengl et al.

(2016). The pink represents the scFv that has been attached to monospecific antibody.

Therapeutic effect

One IgG-scFv that have reached phase I in clinical trials is generated by the company Eli Lilly. This antibody is developed to target HER1 + cMET in solid tumors. (Brinkmann &

Kontermann 2017) Production method

The production of these formats is, because of their symmetric design, easy and standard processes used in the production of antibody fragments and IgG-like antibodies can be applied. Because they contain both Fc region and scFvs, it is assumed that protein A, G and protein L can be used.

Advantages and disadvantages

As noted, the advantage of using fragments like scFv, instead of the original Fab arms, is that mispairing between light chains and/or heavy chains can more easily be avoided. However, many critical issues relating to the stability of the linker between the IgG and the appended scFv fragment have been experienced in antibodies of this kind. Though, existing formats have been developed to improve the structure of the appended IgGs to solve the problems, for example using different linkers. (Brinkmann & Kontermann 2017, Sedykh et al. 2018)

4.1.2.2 scFv-Fc fusions

The fusion idea was extended to create IgG-like antibodies by fusion of different binding modules to an Fc region, generating a scFv-Fc. (Brinkmann & Kontermann 2017, Sedykh et al. 2018)

Structure

The Dual-Affinity Re-Targeting (DART) is one fragment format that have been used to create bispecific antibodies by fragment-Fc fusions. DART fragments are based on the diabody

(28)

22

format. A diabody consists of two separate polypeptide chains, each with a variable heavy and light chain domain with the binding specificity of an antibody, see figure 4. DART is further developed with additional stabilization by covalently linking the polypeptides through a C- terminal disulfide bridge, see figure 14. Studies have shown large increase in stability of the DART molecules in comparison to other diabodies e.g. BiTEs. (Brinkmann & Kontermann 2017, Rader 2011)

Figure 14. A schematic figure of a DART fragment as presented in figure 2, box 9 in the review by Brinkmann & Kontermann (2017). Fusion of one DART fragment represented in pink and purple to an Fc region.

DART-Fc fusion proteins have primarily been developed for T cell retargeting by containing a CD3 binding site. By combining this with the equipment of different antibody binding sites DART-Fc have proven to successfully target and counteract an array of multiple variants of commonly occurring diseases. These range from colorectal cancer and acute myeloid leukemia (AML) to solid tumors. These applications are currently in clinical phase I and II.

(Brinkmann & Kontermann 2017) Production method

A bispecific antibody of this format can be produced by fusion of two different binding domains, each to a separate Fc chain. Any problems with impurities can then be avoided by applying one of the existing strategies that drives heterodimerization. For example, DART-Fc antibodies can be generated by fusing one or more DART fragments to one Fc that contains the KIH substitutions. This design enables purification with protein A and G. (Brinkmann &

Kontermann 2017)

Though being smaller in size compared to native IgG it is possible to retain any desired Fc effector functions and many of the other properties relating to IgGs. Furthermore, fusing fragments to Fc region prolongs the short half-life of Fc-less antibody fragments.

(29)

23

4.1.3 BEAT bispecific antibodies

The biotech company Glenmark have currently developed three formats based on the BEAT platform described in section 3.2.3. The first bispecific antibody generated from the

proprietary BEAT platform is called GBR 1302. The two other formats are described in appendix 4.

Structure

The structure of the antibody is consistent with the BEAT format and has one CD3 binding scFv portion fused to a BEAT heavy chain and one Fab targeting the HER2 receptor.

(WO2015063339A1, 2015; Chen et al. 2019) Therapeutic effect

GBR 1302 is a bispecific antibody that targets the HER2 and CD3 antigen simultaneously.

This enables treatment of different HER2 positive cancers, such as breast and gastric cancer, by binding the CD3 T cell antigen and the HER2 receptor which is overexpressed in many tumor cells. (WO2015063339A1, 2015)

Production methods

In the patent describing the BEAT antibodies (WO2015063339A1, 2015), all antibodies were produced by using following production steps. Preparing DNA expression vectors using standard molecular technology. This can be made by preparing three expression constructs, one for the scFv-Fc fusion, one for the heavy chain and one expressing the corresponding light chain. Following, the DNA vector(s) are transfected or co-transfected into a mammalian cell line. In the patent, HEK cells were used but also CHO cells can be used according to Glenmark. The BEAT antibodies were purified using a two-step process consisting of a capture-elution mode chromatography step using protein G. This step is followed by gradient mode chromatography using protein A chromatography and pH elution. Examples of possible protein A resins that can be used in the purification is MabSelect SuRe or Mabselect protein A column but are not limited to these two. (WO2015063339A1, 2015)

In a study by Skegro et al. (2017) several BEAT antibodies of different formats were constructed and resulted in up to 95% purity after purification with protein A. According to Glenmark the BEAT antibodies can be produced by using conventional production methods.

One study produced BEAT antibodies using a platform approach consisting of expression in CHO-cells and one-step purification with protein A. This resulted in purification levels of 97% which supports Glenmark's suggestion (Moretti et al. 2013). Heterodimerization also showed higher efficiency when compared to using the KIH and CrossMab approach. (Skegro et al. 2017)

Either stepwise or gradient pH elution can be applied in the elution process. This efficiently dissociate most interaction and the homo and heterodimers usually separate with one pH unit in the range between 3-4. It is important to consider that some antibodies could be damaged

(30)

24

by low pH and immediate neutralization and minimizing the time in low-pH are necessary.

(WO2016071355A1, 2016)

4.1.4 XmAb

The biotech company Xencor has developed antibodies with their platform called XmAb. The platform enables alterations with desirable effects to the Fc domain of the antibodies. The following section describes the XmAb formats, with focus on the bispecific antibody XmAb22841.

Structure

One of the XmAb technologies are the Xtend technology that modifies the Fc domain with two amino acid changes. The modification increases affinity to the neonatal Fc receptor which prevents the antibody from degradation. Hence, this interaction extends the antibody’s half- life, making it a more desirable therapeutic drug. (Xencor, 2019)

The bispecific antibody, XmAb22841, with the Xtend modification recently reached phase I.

Other XmAb bispecific antibodies in clinical phase I are of the same format as XmAb22841.

This format is similar in structure to an IgG molecule, but with one Fab arm replaced by a scFv, see figure 15 (Xencor, 2019).

Figure 15. A schematic figure of an XmAb format as presented in “Simultaneous checkpoint-checkpoint or checkpoint-costimulatory receptor targeting with bispecific antibodies promotes enhanced human T cell activation” by Hedvat et al. (2018). The format is composed of an IgG molecule, but with one Fab arm replaced by a scFv represented. The Xtend modification marked in orange on the Fc region extends the antibody's half- life.

The main therapeutic field of bispecific XmAb formats in preclinical and clinical

development is oncology according to Xencor’s pipeline. In the case of XmAb22841, the Fab arm targets LAG-3, a lymphocyte activation gene that suppresses T cell activation and

(31)

25

cytokine secretion (Long et al. 2018). The scFv targets the antigen CTLA-4, which inhibits immune responses (Buchbinder & Desai 2016).

In a study by Moore et al. (2019), the XmAb platform was engineered and tested in the process of producing a bispecific XmAb antibody. It was described that the platform enables antibodies with a heterodimeric Fc domain, i.e. an Fc domain with two different heavy chains.

The upstream and downstream process of the XmAb antibody was described based on proteins containing a heterodimeric Fc. Due to overall high heterodimer yield of 95% and similar biophysical properties, they claimed that other bispecific XmAbs can be manufactured with the same approach as in the study. The approach described below is based on the same format as XmAb22841.

In order to construct the XmAb format an antibody heavy and light chain and a scFv Fc- fusion were subcloned into vectors. The scFv and Fc region were connected with a GS-linker.

The Fc region was altered with substitutions in order to increase the differences in pI between the two heavy chains. This would increase the pI differences between homodimers and heterodimers, which would then facilitate the purification of heterodimers. They sought to minimize the risk of immunogenicity by utilizing buried substitutions, but the exact risk has to be further investigated in clinical studies. For the production of the proteins, plasmids

encoding all chains were co-transfected into HEK cells. The antibody was purified using protein A chromatography and ion exchange chromatography. (Moore et al. 2019) Advantages and disadvantages

As mentioned, the Xencor’s Xtend technology extends the antibodies half-lives. This is advantageous since less frequent medication is needed. Due to alterations in the Fc domain, pI differences can be used in the purification. Furthermore, as for BEAT described in section 4.1.3, the use of one scFv in the design of XmAb22841 circumvents the light chain mispairing problem. The risk of immunogenicity for the XmAb formats was sought to be minimized in the study by Moore et al. (2019). However, since this was not determined it is still unclear whether the formats are immunogenic or not.

4.2 Fragment-based formats

Several formats that heavily deviate from the native Ig-like structure are in development, as will be seen in this section. The bispecific antibodies that were the first to survive clinical trials and reach the open market are such. These fragment-based formats both have

advantages and disadvantages compared to the Ig-like formats. Since they lack the Fc region, they are not as efficiently purified by protein A and protein G chromatography as the Ig-like antibodies. Antibody fragments are instead usually purified by using a His-tag or protein L chromatography. There are numerous of Fc lacking formats, such as BiTEs, BiKEs, TriKEs and DARTs (Duell et al. 2019). These fragments consist of scFvs linked in different ways.

(32)

26

4.2.1 BiTE

The following section describes the fragment-based format BiTE, which stands for bispecific T cell engager as well as the drug blinatumomab developed by Amgen. Blinatumomab is the first representative of the BiTE class that has been approved for use. (Sedykh et al. 2018) Structure

BiTE is a class of small molecules comprised of two scFvs with different specificities, which are connected via a flexible GS-linker, see figure 16 (Huehls et al. 2015).

Figure 16. A schematic figure of a Bispecific T cell Engager (BiTE) as presented in figure 2, box 3 in a review by Brinkmann & Kontermann (2017). The fragment is composed of two scFvs represented in green and blue. The scFvs have different specificities and are connected with a GS-linker.

BiTE simultaneously targets CD3, a receptor that activates cytotoxic T lymphocytes, and a surface antigen of tumor cells (Sedykh et al. 2018). The cytotoxic activity is only activated when both specificities are targeted, which in turn circumvent the issue of undesired T cell activation (Huehls et al. 2015). Currently, there are only two bsAbs approved for use in the US and Europe; blinatumomab and catumaxomab. blinatumomab belongs to the BiTE class and this therapeutic drug targets the proteins CD3 and CD19, the latter of which is a surface protein of B lymphocytes. Blinatumomab is used for the treatment of acute lymphoblastic leukemia. (Sedykh et al. 2018)

In a study by Naddafi et al. (2018), the upstream and downstream process of blinatumomab was described. Both CHO cells and Escherichia coli (E. coli) strains where tested as

expression hosts for the upstream process. The gene coding for blinatumomab was cloned into expression vectors, 6xHis-tagged and purified on a Ni-NTA chromatography column. The Ni- NTA column is used for purification of 6xHis-tagged recombinant proteins (Thermo Fisher, 2019). The result showed that the purified antibody from the CHO cell expression system had higher binding activity compared with the purified antibody from the E. coli expression system. This is due to a more properly folding of proteins in mammalian cells compared to E.

coli cells.

Drugs of the Future - Bispecific Antibodies

19-X2